Matthew R. Linford¹, Alvaro J. Lizarbe¹, David E. Aspnes², Maks Gerashchenko¹, Tyler Andersen¹, Chander Boss¹, Michael P. Jones^1
1Brigham Young University, Provo, UT
²North Carolina State University, Raleigh, NC
Sometimes it is either not practical or possible to collect high quality (high signal-to-noise) X-ray photoelectron spectroscopy (XPS) data. For example, time on XPS instruments is expensive, there can be a high demand on the time of an instrument, and some samples take a considerable amount of time to analyze (even many hours). In addition, sample damage may occur during an analysis. Thus, it can be advantageous to take the least amount of data possible from a sample. For these reasons, the denoising of XPS data can be advantageous. That is, it may be necessary or advantageous to collect imperfect data and then remove the noise from it. Fourier analysis provides an outstanding way to approach the problem of data denoising. Fourier analysis creates an alternative representation of a spectrum in a reciprocal space where it naturally separates the spectrum into signal-containing and noise-containing Fourier coefficients. Accordingly, the multiplication of the Fourier coefficients of a spectrum by an appropriately placed and appropriately defined filter function can largely remove the noise from a spectrum while preserving its underlying spectral form. In this talk, we discuss the Fourier denoising of XPS data with different filter functions, which can be well tailored to this problem. We also discuss the binary residual map as a useful tool for understanding and optimizing smooths.